Artificial Intelligence for Payer Population Health Risk stratification, care gaps, and intervention prioritization in chronic care

Abstract

Payer population health requires a distinction between predicting an adverse outcome, identifying a documented care gap, and selecting a response. We organized public evidence on computational approaches to these functions in chronic care, with Latin America as the context of interest and international studies retained as a separate comparator layer. A structured directed review used a July 16, 2026 search and October 2 targeted source verification, rather than a comprehensive update. We interpreted peer-reviewed research, other academic reports, and official institutional descriptions according to their tasks, accessible content, and evaluation maturity. International studies demonstrated retrospective modeling with administrative data, while care-sequence analyses and non-AI workflows illustrated different approaches to assessing gaps and organizing follow-up. Internal validation, external cohort validation, prospective utility, intervention effects, and routine implementation were treated separately. Official cases described programs or governance, not independently established AI effectiveness. The retrieved Latin American sources provide methodological and non-AI operational context, not evidence of prospectively validated payer AI or demonstrated intervention benefit. Authors' recommendations emphasize explicit outcome definitions, data and subgroup checks, human response design, and staged evaluation before operational use. External evidence supports an evaluation agenda; it does not establish local transferability, clinical benefit, or cost savings.